Re: writing to hdfs on master node much faster
What machines are HDFS data nodes -- just your master? that would explain it. Otherwise, is it actually the write that's slow or is something else you're doing much faster on the master for other reasons maybe? like you're actually shipping data via the master first in some local computation? so the master's executor has the result much faster? On Mon, Apr 20, 2015 at 12:21 PM, jamborta jambo...@gmail.com wrote: Hi all, I have a three node cluster with identical hardware. I am trying a workflow where it reads data from hdfs, repartitions it and runs a few map operations then writes the results back to hdfs. It looks like that all the computation, including the repartitioning and the maps complete within similar time intervals on all the nodes, except when it writes it back to HDFS when the master node does the job way much faster then the slaves (15s for each block as opposed to 1.2 min for the slaves). Any suggestion what the reason might be? thanks, -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/writing-to-hdfs-on-master-node-much-faster-tp22570.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
Re: writing to hdfs on master node much faster
Not sure what would slow it down as the repartition completes equally fast on all nodes, implying that the data is available on all, then there are a few computation steps none of them local on the master. On Mon, Apr 20, 2015 at 12:57 PM, Sean Owen so...@cloudera.com wrote: What machines are HDFS data nodes -- just your master? that would explain it. Otherwise, is it actually the write that's slow or is something else you're doing much faster on the master for other reasons maybe? like you're actually shipping data via the master first in some local computation? so the master's executor has the result much faster? On Mon, Apr 20, 2015 at 12:21 PM, jamborta jambo...@gmail.com wrote: Hi all, I have a three node cluster with identical hardware. I am trying a workflow where it reads data from hdfs, repartitions it and runs a few map operations then writes the results back to hdfs. It looks like that all the computation, including the repartitioning and the maps complete within similar time intervals on all the nodes, except when it writes it back to HDFS when the master node does the job way much faster then the slaves (15s for each block as opposed to 1.2 min for the slaves). Any suggestion what the reason might be? thanks, -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/writing-to-hdfs-on-master-node-much-faster-tp22570.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
RE: writing to hdfs on master node much faster
Check whether your partitioning results in balanced partitions ie partitions with similar sizes - one of the reasons for the performance differences observed by you may be that after your explicit repartitioning, the partition on your master node is much smaller than the RDD partitions on the other 2 nodes -Original Message- From: Sean Owen [mailto:so...@cloudera.com] Sent: Monday, April 20, 2015 12:57 PM To: jamborta Cc: user@spark.apache.org Subject: Re: writing to hdfs on master node much faster What machines are HDFS data nodes -- just your master? that would explain it. Otherwise, is it actually the write that's slow or is something else you're doing much faster on the master for other reasons maybe? like you're actually shipping data via the master first in some local computation? so the master's executor has the result much faster? On Mon, Apr 20, 2015 at 12:21 PM, jamborta jambo...@gmail.com wrote: Hi all, I have a three node cluster with identical hardware. I am trying a workflow where it reads data from hdfs, repartitions it and runs a few map operations then writes the results back to hdfs. It looks like that all the computation, including the repartitioning and the maps complete within similar time intervals on all the nodes, except when it writes it back to HDFS when the master node does the job way much faster then the slaves (15s for each block as opposed to 1.2 min for the slaves). Any suggestion what the reason might be? thanks, -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/writing-to-hdfs-on -master-node-much-faster-tp22570.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
writing to hdfs on master node much faster
Hi all, I have a three node cluster with identical hardware. I am trying a workflow where it reads data from hdfs, repartitions it and runs a few map operations then writes the results back to hdfs. It looks like that all the computation, including the repartitioning and the maps complete within similar time intervals on all the nodes, except when it writes it back to HDFS when the master node does the job way much faster then the slaves (15s for each block as opposed to 1.2 min for the slaves). Any suggestion what the reason might be? thanks, -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/writing-to-hdfs-on-master-node-much-faster-tp22570.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org